faiss 0.1.7 → 0.2.3

data/vendor/faiss/faiss/utils/utils.h

@@ -17,6 +17,7 @@
 #define FAISS_utils_h
 
 #include <stdint.h>
+#include <string>
 
 #ifdef _MSC_VER
 #define strtok_r strtok_s
@@ -24,29 +25,31 @@
 
 #include <faiss/utils/Heap.h>
 
-
 namespace faiss {
 
+/****************************************************************************
+ * Get compile specific variables
+ ***************************************************************************/
+
+/// get compile options
+std::string get_compile_options();
 
 /**************************************************
  * Get some stats about the system
-**************************************************/
-
+ **************************************************/
 
 /// ms elapsed since some arbitrary epoch
-double getmillisecs ();
+double getmillisecs();
 
 /// get current RSS usage in kB
-size_t get_mem_usage_kb ();
+size_t get_mem_usage_kb();
 
-
-uint64_t get_cycles ();
+uint64_t get_cycles();
 
 /***************************************************************************
  * Misc  matrix and vector manipulation functions
  ***************************************************************************/
 
-
 /** compute c := a + bf * b for a, b and c tables
  *
  * @param n   size of the tables
@@ -54,34 +57,37 @@ uint64_t get_cycles ();
  * @param b   size n
  * @param c   restult table, size n
  */
-void fvec_madd (size_t n, const float *a,
-                float bf, const float *b, float *c);
-
+void fvec_madd(size_t n, const float* a, float bf, const float* b, float* c);
 
 /** same as fvec_madd, also return index of the min of the result table
  * @return    index of the min of table c
  */
-int fvec_madd_and_argmin (size_t n, const float *a,
-                           float bf, const float *b, float *c);
-
+int fvec_madd_and_argmin(
+        size_t n,
+        const float* a,
+        float bf,
+        const float* b,
+        float* c);
 
 /* perform a reflection (not an efficient implementation, just for test ) */
-void reflection (const float * u, float * x, size_t n, size_t d, size_t nu);
-
+void reflection(const float* u, float* x, size_t n, size_t d, size_t nu);
 
 /** compute the Q of the QR decomposition for m > n
  * @param a   size n * m: input matrix and output Q
  */
-void matrix_qr (int m, int n, float *a);
+void matrix_qr(int m, int n, float* a);
 
 /** distances are supposed to be sorted. Sorts indices with same distance*/
-void ranklist_handle_ties (int k, int64_t *idx, const float *dis);
+void ranklist_handle_ties(int k, int64_t* idx, const float* dis);
 
 /** count the number of comon elements between v1 and v2
  * algorithm = sorting + bissection to avoid double-counting duplicates
  */
-size_t ranklist_intersection_size (size_t k1, const int64_t *v1,
-                                   size_t k2, const int64_t *v2);
+size_t ranklist_intersection_size(
+        size_t k1,
+        const int64_t* v1,
+        size_t k2,
+        const int64_t* v2);
 
 /** merge a result table into another one
  *
@@ -91,41 +97,38 @@ size_t ranklist_intersection_size (size_t k1, const int64_t *v1,
  * @param translation  add this value to all I1's indexes
  * @return             nb of values that were taken from the second table
  */
-size_t merge_result_table_with (size_t n, size_t k,
-                                int64_t *I0, float *D0,
-                                const int64_t *I1, const float *D1,
-                                bool keep_min = true,
-                                int64_t translation = 0);
-
+size_t merge_result_table_with(
+        size_t n,
+        size_t k,
+        int64_t* I0,
+        float* D0,
+        const int64_t* I1,
+        const float* D1,
+        bool keep_min = true,
+        int64_t translation = 0);
 
 /// a balanced assignment has a IF of 1
-double imbalance_factor (int n, int k, const int64_t *assign);
+double imbalance_factor(int n, int k, const int64_t* assign);
 
 /// same, takes a histogram as input
-double imbalance_factor (int k, const int *hist);
-
-
-void fvec_argsort (size_t n, const float *vals,
-                    size_t *perm);
+double imbalance_factor(int k, const int* hist);
 
-void fvec_argsort_parallel (size_t n, const float *vals,
-                    size_t *perm);
+void fvec_argsort(size_t n, const float* vals, size_t* perm);
 
+void fvec_argsort_parallel(size_t n, const float* vals, size_t* perm);
 
 /// compute histogram on v
-int ivec_hist (size_t n, const int * v, int vmax, int *hist);
+int ivec_hist(size_t n, const int* v, int vmax, int* hist);
 
 /** Compute histogram of bits on a code array
  *
  * @param codes   size(n, nbits / 8)
  * @param hist    size(nbits): nb of 1s in the array of codes
  */
-void bincode_hist(size_t n, size_t nbits, const uint8_t *codes, int *hist);
-
+void bincode_hist(size_t n, size_t nbits, const uint8_t* codes, int* hist);
 
 /// compute a checksum on a table.
-size_t ivec_checksum (size_t n, const int *a);
-
+size_t ivec_checksum(size_t n, const int* a);
 
 /** random subsamples a set of vectors if there are too many of them
  *
@@ -136,9 +139,13 @@ size_t ivec_checksum (size_t n, const int *a);
  * @param seed   random seed to use for sampling
  * @return       x or an array allocated with new [] with *n vectors
  */
-const float *fvecs_maybe_subsample (
-       size_t d, size_t *n, size_t nmax, const float *x,
-       bool verbose = false, int64_t seed = 1234);
+const float* fvecs_maybe_subsample(
+        size_t d,
+        size_t* n,
+        size_t nmax,
+        const float* x,
+        bool verbose = false,
+        int64_t seed = 1234);
 
 /** Convert binary vector to +1/-1 valued float vector.
  *
@@ -146,7 +153,7 @@ const float *fvecs_maybe_subsample (
  * @param x_in   input binary vector (uint8_t table of size d / 8)
  * @param x_out  output float vector (float table of size d)
  */
-void binary_to_real(size_t d, const uint8_t *x_in, float *x_out);
+void binary_to_real(size_t d, const uint8_t* x_in, float* x_out);
 
 /** Convert float vector to binary vector. Components > 0 are converted to 1,
  * others to 0.
@@ -155,16 +162,14 @@ void binary_to_real(size_t d, const uint8_t *x_in, float *x_out);
  * @param x_in   input float vector (float table of size d)
  * @param x_out  output binary vector (uint8_t table of size d / 8)
  */
-void real_to_binary(size_t d, const float *x_in, uint8_t *x_out);
-
+void real_to_binary(size_t d, const float* x_in, uint8_t* x_out);
 
 /** A reasonable hashing function */
-uint64_t hash_bytes (const uint8_t *bytes, int64_t n);
+uint64_t hash_bytes(const uint8_t* bytes, int64_t n);
 
 /** Whether OpenMP annotations were respected. */
 bool check_openmp();
 
-} // namspace faiss
-
+} // namespace faiss
 
 #endif /* FAISS_utils_h */

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: faiss
 version: !ruby/object:Gem::Version
-  version: 0.1.7
+  version: 0.2.3
 platform: ruby
 authors:
 - Andrew Kane
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2021-03-26 00:00:00.000000000 Z
+date: 2021-12-17 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: rice
@@ -16,14 +16,14 @@ dependencies:
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: '2.2'
+        version: 4.0.2
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: '2.2'
+        version: 4.0.2
 - !ruby/object:Gem::Dependency
   name: numo-narray
   requirement: !ruby/object:Gem::Requirement
@@ -53,16 +53,12 @@ files:
 - ext/faiss/index.cpp
 - ext/faiss/index_binary.cpp
 - ext/faiss/kmeans.cpp
+- ext/faiss/numo.hpp
 - ext/faiss/pca_matrix.cpp
 - ext/faiss/product_quantizer.cpp
 - ext/faiss/utils.cpp
 - ext/faiss/utils.h
 - lib/faiss.rb
-- lib/faiss/index.rb
-- lib/faiss/index_binary.rb
-- lib/faiss/kmeans.rb
-- lib/faiss/pca_matrix.rb
-- lib/faiss/product_quantizer.rb
 - lib/faiss/version.rb
 - vendor/faiss/LICENSE
 - vendor/faiss/faiss/AutoTune.cpp
@@ -107,6 +103,10 @@ files:
 - vendor/faiss/faiss/IndexLSH.h
 - vendor/faiss/faiss/IndexLattice.cpp
 - vendor/faiss/faiss/IndexLattice.h
+- vendor/faiss/faiss/IndexNNDescent.cpp
+- vendor/faiss/faiss/IndexNNDescent.h
+- vendor/faiss/faiss/IndexNSG.cpp
+- vendor/faiss/faiss/IndexNSG.h
 - vendor/faiss/faiss/IndexPQ.cpp
 - vendor/faiss/faiss/IndexPQ.h
 - vendor/faiss/faiss/IndexPQFastScan.cpp
@@ -117,6 +117,8 @@ files:
 - vendor/faiss/faiss/IndexRefine.h
 - vendor/faiss/faiss/IndexReplicas.cpp
 - vendor/faiss/faiss/IndexReplicas.h
+- vendor/faiss/faiss/IndexResidual.cpp
+- vendor/faiss/faiss/IndexResidual.h
 - vendor/faiss/faiss/IndexScalarQuantizer.cpp
 - vendor/faiss/faiss/IndexScalarQuantizer.h
 - vendor/faiss/faiss/IndexShards.cpp
@@ -175,6 +177,8 @@ files:
 - vendor/faiss/faiss/gpu/utils/StaticUtils.h
 - vendor/faiss/faiss/gpu/utils/Timer.cpp
 - vendor/faiss/faiss/gpu/utils/Timer.h
+- vendor/faiss/faiss/impl/AdditiveQuantizer.cpp
+- vendor/faiss/faiss/impl/AdditiveQuantizer.h
 - vendor/faiss/faiss/impl/AuxIndexStructures.cpp
 - vendor/faiss/faiss/impl/AuxIndexStructures.h
 - vendor/faiss/faiss/impl/FaissAssert.h
@@ -182,11 +186,19 @@ files:
 - vendor/faiss/faiss/impl/FaissException.h
 - vendor/faiss/faiss/impl/HNSW.cpp
 - vendor/faiss/faiss/impl/HNSW.h
+- vendor/faiss/faiss/impl/LocalSearchQuantizer.cpp
+- vendor/faiss/faiss/impl/LocalSearchQuantizer.h
+- vendor/faiss/faiss/impl/NNDescent.cpp
+- vendor/faiss/faiss/impl/NNDescent.h
+- vendor/faiss/faiss/impl/NSG.cpp
+- vendor/faiss/faiss/impl/NSG.h
 - vendor/faiss/faiss/impl/PolysemousTraining.cpp
 - vendor/faiss/faiss/impl/PolysemousTraining.h
 - vendor/faiss/faiss/impl/ProductQuantizer-inl.h
 - vendor/faiss/faiss/impl/ProductQuantizer.cpp
 - vendor/faiss/faiss/impl/ProductQuantizer.h
+- vendor/faiss/faiss/impl/ResidualQuantizer.cpp
+- vendor/faiss/faiss/impl/ResidualQuantizer.h
 - vendor/faiss/faiss/impl/ResultHandler.h
 - vendor/faiss/faiss/impl/ScalarQuantizer.cpp
 - vendor/faiss/faiss/impl/ScalarQuantizer.h
@@ -228,6 +240,7 @@ files:
 - vendor/faiss/faiss/utils/distances.cpp
 - vendor/faiss/faiss/utils/distances.h
 - vendor/faiss/faiss/utils/distances_simd.cpp
+- vendor/faiss/faiss/utils/extra_distances-inl.h
 - vendor/faiss/faiss/utils/extra_distances.cpp
 - vendor/faiss/faiss/utils/extra_distances.h
 - vendor/faiss/faiss/utils/hamming-inl.h
@@ -243,9 +256,10 @@ files:
 - vendor/faiss/faiss/utils/simdlib.h
 - vendor/faiss/faiss/utils/simdlib_avx2.h
 - vendor/faiss/faiss/utils/simdlib_emulated.h
+- vendor/faiss/faiss/utils/simdlib_neon.h
 - vendor/faiss/faiss/utils/utils.cpp
 - vendor/faiss/faiss/utils/utils.h
-homepage: https://github.com/ankane/faiss
+homepage: https://github.com/ankane/faiss-ruby
 licenses:
 - MIT
 metadata: {}
@@ -257,14 +271,14 @@ required_ruby_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
-      version: '2.4'
+      version: '2.6'
 required_rubygems_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.2.3
+rubygems_version: 3.2.32
 signing_key:
 specification_version: 4
 summary: Efficient similarity search and clustering for Ruby

data/lib/faiss/index.rb

@@ -1,20 +0,0 @@
-module Faiss
-  class Index
-    def train(objects)
-      objects = Numo::SFloat.cast(objects) unless objects.is_a?(Numo::SFloat)
-      _train(objects.shape[0], objects)
-    end
-
-    def add(objects)
-      objects = Numo::SFloat.cast(objects) unless objects.is_a?(Numo::SFloat)
-      _add(objects.shape[0], objects)
-    end
-
-    def search(objects, k)
-      objects = Numo::SFloat.cast(objects) unless objects.is_a?(Numo::SFloat)
-      n = objects.shape[0]
-      distances, labels = _search(n, objects, k)
-      [Numo::SFloat.from_binary(distances).reshape(n, k), Numo::Int64.from_binary(labels).reshape(n, k)]
-    end
-  end
-end

data/lib/faiss/index_binary.rb

@@ -1,20 +0,0 @@
-module Faiss
-  class IndexBinary
-    def train(objects)
-      objects = Numo::UInt8.cast(objects) unless objects.is_a?(Numo::UInt8)
-      _train(objects.shape[0], objects)
-    end
-
-    def add(objects)
-      objects = Numo::UInt8.cast(objects) unless objects.is_a?(Numo::UInt8)
-      _add(objects.shape[0], objects)
-    end
-
-    def search(objects, k)
-      objects = Numo::UInt8.cast(objects) unless objects.is_a?(Numo::UInt8)
-      n = objects.shape[0]
-      distances, labels = _search(n, objects, k)
-      [Numo::UInt32.from_binary(distances).reshape(n, k), Numo::Int64.from_binary(labels).reshape(n, k)]
-    end
-  end
-end

data/lib/faiss/kmeans.rb

@@ -1,15 +0,0 @@
-module Faiss
-  class Kmeans
-    attr_reader :index
-
-    def train(objects)
-      objects = Numo::SFloat.cast(objects) unless objects.is_a?(Numo::SFloat)
-      @index = IndexFlatL2.new(d)
-      _train(objects.shape[0], objects, @index)
-    end
-
-    def centroids
-      Numo::SFloat.from_binary(_centroids).reshape(k, d)
-    end
-  end
-end

data/lib/faiss/pca_matrix.rb

@@ -1,15 +0,0 @@
-module Faiss
-  class PCAMatrix
-    def train(objects)
-      objects = Numo::SFloat.cast(objects) unless objects.is_a?(Numo::SFloat)
-      _train(objects.shape[0], objects)
-    end
-
-    def apply(objects)
-      objects = Numo::SFloat.cast(objects) unless objects.is_a?(Numo::SFloat)
-      n = objects.shape[0]
-      res = _apply(n, objects)
-      Numo::SFloat.from_binary(res).reshape(n, d_out)
-    end
-  end
-end

data/lib/faiss/product_quantizer.rb

@@ -1,22 +0,0 @@
-module Faiss
-  class ProductQuantizer
-    def train(objects)
-      objects = Numo::SFloat.cast(objects) unless objects.is_a?(Numo::SFloat)
-      _train(objects.shape[0], objects)
-    end
-
-    def compute_codes(objects)
-      objects = Numo::SFloat.cast(objects) unless objects.is_a?(Numo::SFloat)
-      n = objects.shape[0]
-      res = _compute_codes(n, objects)
-      Numo::UInt8.from_binary(res).reshape(n, m)
-    end
-
-    def decode(objects)
-      objects = Numo::UInt8.cast(objects) unless objects.is_a?(Numo::UInt8)
-      n = objects.shape[0]
-      res = _decode(n, objects)
-      Numo::SFloat.from_binary(res).reshape(n, d)
-    end
-  end
-end